Spatiotemporal tensor analysis for effective information mining of hydraulic structures considering environmental excitation and vibration response.

The vibration response data is a key foundation of vibration-based hydraulic structures' online damage diagnosis. However, the measured data is often subject to various noises and invalid information, which reduces the accuracy of damage diagnosis, leading to misjudgment and omission of structure damage. The hydraulic structure system is an open, dissipative, and complex nonlinear dynamic system, where at least one or more, or even all parts, have nonlinear interactions. The service condition of hydraulic concrete structures is influenced by environmental factors such as temperature, water temperature and water level. The feature of "open" is mainly manifested as the coupling effect field of multiphase environmental factors. The single-point signal denoising based effective information mining method can lead to over-denoising or under-denoising issues, resulting in low effective information mining accuracy. To overcome these limitations, this paper studies the synchronous denoising technology of multi-point vibration response data, and an improved adaptive variational mode decomposition method was introduced to convert the multi-point vibration response data into a three-dimensional multi-scale spatiotemporal tensor. The factor filter sets construction method based on the Crank-Nicolson-like criterion for fast orthogonal Tucker factor updating method was proposed to denoise the signal preliminary. The time-weighted modified dynamic time warping theory and curvature smoothing algorithm were combined to construct the optimal filter model with a balancing factor to extract the effective information from vibration response. Finally, the data from the sluice model experiment was used to demonstrate the validity of the method proposed in this paper. This article is about the theme of health monitoring for hydraulic concrete structures in the field of civil engineering.